1 TADPOLE and BSWiMS

1.0.1 Loading the libraries

library("FRESA.CAD")
library(survival)
library(readxl)
library(igraph)
op <- par(no.readonly = TRUE)
pander::panderOptions('digits', 3)
pander::panderOptions('table.split.table', 400)
pander::panderOptions('keep.trailing.zeros',TRUE)

1.1 Loading BSWiMS Results


load("./TADPOLE_BSWIMS_Results.RData")

pander::pander(table(TADPOLE_Conv_TRAIN$status))

0	1
261	133

pander::pander(table(TADPOLE_Conv_TEST$status))

0	1
112	58

par(op)

1.1.1 Cross Validation


cvLASSORaw <- randomCV(TADPOLECrossMRI,
                 Surv(TimeToEvent,status)~.,
                 fittingFunction= LASSO_1SE,
                 trainSampleSets= cvBSWiMSRaw$trainSamplesSets,
 )

……….10 Tested: 552 Avg. Selected: 8.8 Min Tests: 1 Max Tests: 10 Mean Tests: 5.036232 . MAD: 1.847167 ……….20 Tested: 562 Avg. Selected: 9.05 Min Tests: 1 Max Tests: 17 Mean Tests: 9.893238 . MAD: 2.368578 ……….30 Tested: 564 Avg. Selected: 8.7 Min Tests: 3 Max Tests: 25 Mean Tests: 14.78723 . MAD: 2.34703 ……….40 Tested: 564 Avg. Selected: 8.95 Min Tests: 4 Max Tests: 32 Mean Tests: 19.71631 . MAD: 2.622609 ……….50 Tested: 564 Avg. Selected: 9.16 Min Tests: 6 Max Tests: 39 Mean Tests: 24.64539 . MAD: 2.539005

pander::pander(cbind(cvLASSORaw$featureFrequency[cvLASSORaw$featureFrequency>20]))

ADAS13	50
FAQ	50
RAVLT_immediate	49
ABETA	43
TAU	28
M_ST32CV	27
M_ST24CV	24
M_ST29SV	23


prBin <- predictionStats_binary(cvLASSORaw$survMedianTrain[,c(2,3)],"TRAIN: MCI to  AD Conversion")


survmtest <- cvLASSORaw$survMedianTest
survmtest <- survmtest[complete.cases(survmtest),]

prBin <- predictionStats_binary(survmtest[,c(2,3)],"LASSO: MCI to  AD Conversion")

pander::pander(prBin$aucs)

est	lower	upper
0.875	0.847	0.904

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	158	84	242
Test -	33	289	322
Total	191	373	564


par(op)
ho <- mean(survmtest$Outcome)
timeInterval <- mean(survmtest[survmtest$Outcome==0,"Times"])
pgzero <- ppoisGzero(survmtest$LinearPredictorsMedian,ho)
rsdata <- cbind(survmtest$Outcome,pgzero,survmtest$Times)
riskAnalysis <- RRPlot(rsdata,riskTimeInterval=timeInterval,title="LASSO")

[1] 0.3386525 [1] 0.3386525 1.0000000 [1] 0.9906981 0.9625794 0.3942505 3.4912148 13.9578736 29.1869870 0.0000000 [8] 1.0000000


pander::pander(riskAnalysis$c.index)

C Index: 0.846
Dxy: 0.693
S.D.: 0.0246
n: 564
missing: 0
uncensored: 191
Relevant Pairs: 142910
Concordant: 120960
Uncertain: 174472
cstatCI:

mean.C Index median lower upper

0.846 0.846 0.82 0.869

mean.C Index	median	lower	upper
0.846	0.846	0.82	0.869

pander::pander(riskAnalysis$ROCAnalysis$aucs)

est	lower	upper
0.875	0.847	0.904

pander::pander(riskAnalysis$cenAUC)

0.897

pander::pander(riskAnalysis$ROCAnalysis$ClassMetrics)

accci:

50% 2.5% 97.5%

0.789 0.754 0.823
senci:

50% 2.5% 97.5%

0.64 0.57 0.708
speci:

50% 2.5% 97.5%

0.866 0.83 0.9
aucci:

50% 2.5% 97.5%

0.753 0.714 0.791
berci:

50% 2.5% 97.5%

0.247 0.209 0.286
preci:

50% 2.5% 97.5%

0.71 0.636 0.774
F1ci:

50% 2.5% 97.5%

0.672 0.614 0.727

50%	2.5%	97.5%
0.789	0.754	0.823

50%	2.5%	97.5%
0.64	0.57	0.708

50%	2.5%	97.5%
0.866	0.83	0.9

50%	2.5%	97.5%
0.753	0.714	0.791

50%	2.5%	97.5%
0.247	0.209	0.286

50%	2.5%	97.5%
0.71	0.636	0.774

50%	2.5%	97.5%
0.672	0.614	0.727

pander::pander(riskAnalysis$surdif)

Call: survival::Surv(eTime, eStatus) ~ class Chisq = 293.389809 on 2 degrees of freedom, p = 0.000000
	N	Observed	Expected	(O-E)^2/E	(O-E)^2/V
class=0	311	28	127.6	77.72	242.05
class=1	81	41	26.5	7.93	9.28
class=2	172	122	36.9	196.05	252.21

1.1.2 Cross Validation RIDGE


cvRIDGERaw <- randomCV(TADPOLECrossMRI,
                 Surv(TimeToEvent,status)~.,
                 fittingFunction= GLMNET_RIDGE_1SE,
                 trainSampleSets= cvBSWiMSRaw$trainSamplesSets,
 )

……….10 Tested: 552 Avg. Selected: 275.6 Min Tests: 1 Max Tests: 10 Mean Tests: 5.036232 . MAD: 7.812309 ……….20 Tested: 562 Avg. Selected: 277.8 Min Tests: 1 Max Tests: 17 Mean Tests: 9.893238 . MAD: 8.190561 ……….30 Tested: 564 Avg. Selected: 276.2667 Min Tests: 3 Max Tests: 25 Mean Tests: 14.78723 . MAD: 8.211094 ……….40 Tested: 564 Avg. Selected: 276.1 Min Tests: 4 Max Tests: 32 Mean Tests: 19.71631 . MAD: 8.308367 ……….50 Tested: 564 Avg. Selected: 276.16 Min Tests: 6 Max Tests: 39 Mean Tests: 24.64539 . MAD: 8.364073


prBin <- predictionStats_binary(cvRIDGERaw$survMedianTrain[,c(2,3)],"TRAIN: MCI to  AD Conversion")


survmtest <- cvRIDGERaw$survMedianTest
survmtest <- survmtest[complete.cases(survmtest),]

prBin <- predictionStats_binary(survmtest[,c(2,3)],"RIDGE: MCI to  AD Conversion")

pander::pander(prBin$aucs)

est	lower	upper
0.591	0.537	0.645

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	114	189	303
Test -	77	184	261
Total	191	373	564


par(op)
ho <- mean(survmtest$Outcome)
timeInterval <- mean(survmtest[survmtest$Outcome==0,"Times"])
pgzero <- ppoisGzero(survmtest$LinearPredictorsMedian,ho)
rsdata <- cbind(survmtest$Outcome,pgzero,survmtest$Times)
riskAnalysis <- RRPlot(rsdata,riskTimeInterval=timeInterval,title="RIDGE")

[1] 0.3386525 [1] 0.3386525 1.0000000 [1] 0.5946290 0.4987360 0.3942505 3.4912148 186.9714797 107.1554930 [7] 0.0000000 0.0000000


pander::pander(riskAnalysis$c.index)

C Index: 0.572
Dxy: 0.144
S.D.: 0.0422
n: 564
missing: 0
uncensored: 191
Relevant Pairs: 142910
Concordant: 81723
Uncertain: 174472
cstatCI:

mean.C Index median lower upper

0.572 0.571 0.531 0.614

mean.C Index	median	lower	upper
0.572	0.571	0.531	0.614

pander::pander(riskAnalysis$ROCAnalysis$aucs)

est	lower	upper
0.535	0.485	0.584

pander::pander(riskAnalysis$cenAUC)

0.54

pander::pander(riskAnalysis$ROCAnalysis$ClassMetrics)

accci:

50% 2.5% 97.5%

0.564 0.523 0.603
senci:

50% 2.5% 97.5%

0.479 0.411 0.552
speci:

50% 2.5% 97.5%

0.609 0.558 0.659
aucci:

50% 2.5% 97.5%

0.544 0.5 0.585
berci:

50% 2.5% 97.5%

0.456 0.415 0.5
preci:

50% 2.5% 97.5%

0.386 0.324 0.446
F1ci:

50% 2.5% 97.5%

0.427 0.368 0.483

50%	2.5%	97.5%
0.564	0.523	0.603

50%	2.5%	97.5%
0.479	0.411	0.552

50%	2.5%	97.5%
0.609	0.558	0.659

50%	2.5%	97.5%
0.544	0.5	0.585

50%	2.5%	97.5%
0.456	0.415	0.5

50%	2.5%	97.5%
0.386	0.324	0.446

50%	2.5%	97.5%
0.427	0.368	0.483

pander::pander(riskAnalysis$surdif)

Call: survival::Surv(eTime, eStatus) ~ class Chisq = 8.796285 on 1 degrees of freedom, p = 0.003018
	N	Observed	Expected	(O-E)^2/E	(O-E)^2/V
class=0	327	100	119.7	3.25	8.8
class=1	237	91	71.3	5.46	8.8

1.1.3 Learning

bConvmlLZO <- LASSO_1SE(Surv(TimeToEvent,status)~.,TADPOLE_Conv_TRAIN)

pander::pander(bConvmlLZO$selectedfeatures)

ADAS13, MMSE, RAVLT_immediate, RAVLT_perc_forgetting, FAQ, APOE4, ABETA, PTAU, M_ST13TA, M_ST56SA, M_ST24CV, M_ST32CV, M_ST40CV, M_ST29SV and RD_ST31TA


ptestl <- predict(bConvmlLZO,TADPOLE_Conv_TEST)
cval <- mean(ptestl)
ptestl <- predict(bConvmlLZO,TADPOLE_Conv_TEST) - cval

boxplot(ptestl~TADPOLE_Conv_TEST$status)

ptestr <- exp(ptestl)

predsurv <- cbind(TADPOLE_Conv_TEST$TimeToEvent,
                  TADPOLE_Conv_TEST$status,
                  ptestl,
                  ptestr)

prSurv <- predictionStats_survival(predsurv,"MCI to  AD Conversion")

pander::pander(prSurv$CIRisk)

median	lower	upper
0.889	0.852	0.92

pander::pander(prSurv$CILp)

median	lower	upper
0.896	0.845	0.939

pander::pander(prSurv$spearmanCI)

50%	2.5%	97.5%
0.568	0.37	0.723


prBin <- predictionStats_binary(cbind(TADPOLE_Conv_TEST$status,ptestl),"MCI to  AD Conversion")

pander::pander(prBin$aucs)

est	lower	upper
0.896	0.85	0.943

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	54	26	80
Test -	4	86	90
Total	58	112	170


par(op)

1.1.4 Learning RIDGE

bConvmlRIDGE <- GLMNET_RIDGE_1SE(Surv(TimeToEvent,status)~.,TADPOLE_Conv_TRAIN)

ptestl <- predict(bConvmlRIDGE,TADPOLE_Conv_TEST)
cval <- mean(ptestl)
ptestl <- predict(bConvmlRIDGE,TADPOLE_Conv_TEST) - cval

boxplot(ptestl~TADPOLE_Conv_TEST$status)

ptestr <- exp(ptestl)

predsurv <- cbind(TADPOLE_Conv_TEST$TimeToEvent,
                  TADPOLE_Conv_TEST$status,
                  ptestl,
                  ptestr)

prSurv <- predictionStats_survival(predsurv,"MCI to  AD Conversion")

pander::pander(prSurv$CIRisk)

median	lower	upper
0.84	0.795	0.882

pander::pander(prSurv$CILp)

median	lower	upper
0.86	0.806	0.91

pander::pander(prSurv$spearmanCI)

50%	2.5%	97.5%
0.355	0.0891	0.581


prBin <- predictionStats_binary(cbind(TADPOLE_Conv_TEST$status,ptestl),"MCI to  AD Conversion")

pander::pander(prBin$aucs)

est	lower	upper
0.86	0.805	0.914

pander::pander(prBin$CM.analysis$tab)

	Outcome +	Outcome -	Total
Test +	50	35	85
Test -	8	77	85
Total	58	112	170


par(op)

1.1.5 Saving the enviroment

save.image("./TADPOLE_LASSO_Results.RData")

TADPOLE_GLMNET

Jose Tamez

2024-10-18